Seasonal precipitation totals and mean air temperatures from the 1997 El Niño–Southern Oscillation (ENSO) warm event are evaluated in light of nine historical warm events (and their nine-member mean) occurring from 1951 through 1987. Comparisons between the 1997 and historical warm events are made for 177 precipitation and 191 temperature records within the continental United States for the winter and spring seasons immediately following the onset of the warm phase. The winter and spring of 1998 contain the peak impacts for the 1997 event and encompass the mature and decay stages of this ENSO warm phase.
The authors’ goal is to provide a historical record of the winter and spring anomalies associated with the 1997 warm event. We hope to show that the 1997 warm phase was associated with anomalies both similar to and unique from historical warm events. In addition, we outline some of the potential causes for event to event variability in the ENSO response over the United States. Pattern shifts in seasonal total precipitation and mean temperature are identified through comparison of the 1998 anomalies to the nine-member composite winter and spring anomalies. Ranking the anomalies versus nine historical warm phases identifies extreme 1998 seasonal anomalies and places these anomalies in a historical context.
The warm phase of ENSO has been shown by many authors (e.g., Ropelewski and Halpert 1986) to significantly impact both the temperature and precipitation patterns over the United States. The warm phase typically brings heavy rains to the Gulf Coast states and drier weather to the Pacific Northwest. Additionally, warmer winters are expected for the northern plains and Midwest. The warm phase is also associated with highly variable weather in some locations. For example, some warm phases bring heavy rains to southern California, while still others result in drier conditions. Herein we examine the temperature and precipitation anomalies for the 1997 warm phase and identify occurrences of both expected and unexpected responses to a warm phase.
Precipitation and temperature anomalies for the winter of 1998 generally correspond to the nine-member composite pattern but with exceptional amplitudes. Winter precipitation at many stations in the southern third of the nation exceeded all nine of the historical warm events, including the 1982 event, with especially surprising increases along the East Coast as far north as Washington, D.C., and New York City. In contrast to the winter, 1998 spring temperature and precipitation anomalies are markedly different from the nine-member composite. In fact, precipitation anomalies in the Pacific Northwest, Texas, and the Gulf Coast during the spring of 1998 are in opposition to the composite patterns.
2. Data and analysis
Monthly historical data for the period 1951–87 are extracted from the Global Historical Climatology Network (GHCN) dataset (Vose et al. 1992). These data have the advantage of being adjusted to remove spurious trends and jumps. For 1997 and 1998, daily precipitation totals and temperature means are obtained from the summary of the day (SOD) data available through the National Climatic Data Center. The 37-yr period from 1951 through 1987 is chosen to maximize the number of stations with both SOD data (<20% missing daily values) and GHCN data with a uniform record length. A uniform record length in the GHCN data limits the influence of instrumental decadal variability in the ENSO phase composites. Our strict matching requirements limited our analysis to ∼10% of the SOD stations: a total of 177 stations for precipitation and 191 stations for temperature. At each matched station, seasonal mean temperature and total precipitation are created for the winter (December–February; DJF) and spring (March–May; MAM) using both the SOD and historical data.
Historical data are categorized into ENSO phases based on the Japanese Meteorological Agency (JMA) sea surface temperature (SST) index (Green 1996). The JMA SST index is a five-month running mean of SST anomalies over the equatorial Pacific Ocean from 4°N to 4°S and from 150° to 90°W. An ENSO warm phase is defined when the JMA SST index is greater than 0.5°C for at least six consecutive months. The consecutive months must begin prior to October and include October, November, and December. A cold phase of ENSO is similarly defined, except that the JMA SST must be less than −0.5°C.
Extremes in the ENSO cycle typically develop during summer, peak in late fall, and decay into the following spring. Therefore we define an ENSO year as the period from October of the onset year through September of the following year. As an example, the 1997 warm phase is defined as the year from October 1997 through September 1998. All years not classified as warm or cold according to the criterion listed above are considered neutral years. Table 1 lists the warm and neutral years used in this study.
A comparison of warm phase years selected using the JMA SST to those selected by Kiladis and Diaz (1989) and Shabbar et al. (1997) reveal minor differences before 1965. Kiladis and Diaz (1989) defined 1953 as a warm year using a combination of a Southern Oscillation index (SOI) and equatorial sea temperatures. Shabbar et al. (1997) used the SOI to define only moderate and strong warm phases, including 1953 and 1958 but excluding the warm phase of 1963. The authors wish to note that no single index of ENSO warm phases is definitive. These indices are continually undergoing refinement (Trenberth 1997) and have been extended into the previous century (Können et al. 1998). Although there is some disagreement, using the JMA SST index has little impact on our results because the majority of the warm phases used in this study are identified by other indices.
Historical ENSO warm and neutral phase composite temperature means and precipitation totals are created by averaging GHCN seasonal means (totals) over the nine warm and 18 neutral phases listed in Table 1. We compare these historically based composites to one another (composite warm minus neutral anomalies) as well as the 1997 warm and GHCN composite neutral (hereafter 97N anomalies), and the 1997 warm and GHCN composite warm (hereafter 97W anomalies) phases. These comparisons facilitate the contrasting of the 1997 warm event to the typical warm and neutral events. All precipitation anomalies are evaluated as percentage differences to account for variations in rainfall at individual stations (e.g., a +100 mm anomaly has a greater impact at a station typically receiving 10 mm of rain than at a station receiving 200 mm of precipitation).
Stations with robust temperature differences are identified for all cases. A Student’s t-test identifies robust differences between GHCN composite warm and neutral phase temperature means. Extreme 97N (97W) temperature anomalies are statistically qualified by determining whether the winter and spring temperature means following the onset of the 1997 warm phase occur in the 5th or 95th percentile of the Gaussian distributions for the historical neutral (warm) phase seasons. In addition, the magnitudes of the 97N temperature anomalies are ranked with respect to the nine historical warm events to highlight regions with remarkable 1997 ENSO impacts.
Precipitation data tend to be non-Gaussian, especially in arid regions, so modified gamma distributions (Wilks 1990) are fit to the historical seasonal precipitation data to assess significance (Wilks 1995). The fit of a gamma distribution to the warm and neutral phase seasonal precipitation at each station is assessed using a Kolmogorov–Smirnov (KS) test (Hollander and Wolfe 1999). Precipitation statistics are only evaluated for stations where the p value of the KS test is equal to or exceeds 0.95 (i.e., the fit matches the data). For the GHCN composite precipitation anomalies, we identify significant shifts between the median of the fitted warm and neutral phase gamma distributions. Gamma probabilities are calculated to identify where the winter and spring precipitation totals following the onset of the 1997 warm phase fall on the fitted warm and neutral phase distributions. Finally, the magnitude of the 97N anomalies are ranked versus the warm minus neutral anomalies for nine historical warm phases to highlight extreme events.
3. 1998 precipitation anomalies
The canonical pattern of warm phase winter precipitation anomalies over the continental United States has been evaluated by several authors. The occurrence of an increase in precipitation along the Gulf Coast has been noted by Ropelewski and Halpert (1996), Green (1996), Sittel (1994a,b), and Kiladis and Diaz (1989). Along the Pacific coastline Smith et al. (1998), Green (1996), and Yarnal and Diaz (1986) found the northwestern states to be drier and the coastline of central California to have increased precipitation. Over southern California Smith et al. (1998) and Sittel (1994b) revealed weak dry anomalies. In addition, Shabbar et al. (1997) found the dry anomalies of the Pacific Northwest to extend inland over southern Canada as far as the western Great Lakes.
Herein our GHCN composite winter anomalies (Fig. 1a) reveal some of the features outlined by these past studies. Statistically significant increases in the median of the warm phase precipitation distribution occur along the Gulf Coast with local anomaly maxima in south Texas and Florida. A further significant shift toward increased precipitation in the mid-Atlantic states also occurs in the GHCN composite. A similar mid-Atlantic pattern was noted by Ropelewski and Halpert (1996), but the pattern did not meet their rigid requirements for signal consistency over their century-long data record. The dry conditions noted by Shabbar et al. (1997) in southern Canada are shown to extend south of the border into the northern Rocky Mountain states and the Great Lakes region; however, only two stations (Denver, Colorado, and Des Moines, Iowa) have a significant median shift toward less precipitation. Along the West Coast the GHCN composite shows only a weak pattern of dry Pacific Northwest, wet central California, and dry southern California noted by other authors.
The pattern of precipitation anomalies during the winter of 1998 (D97, JF98) is similar to the GHCN historical composite, with the notable exception of California. In the GHCN composite (Fig. 1a), percent changes in precipitation over California are small, while the 97W winter precipitation anomalies (Fig. 1c) reveal that many stations in California received over 150% more precipitation than expected for a warm phase. The majority of this precipitation occurred in February 1998 when five stations in the Southwest received over a 300% increase in their monthly total precipitation. For the winter of 1998, most California precipitation totals are in the 95th percentile of both the warm and neutral phase gamma distributions, and seven of the 97N positive precipitation anomalies over California and Nevada rank as the largest of the past 10 warm events (Fig. 2).
One explanation for the precipitation anomalies over California in the winter of 1998 not agreeing with the GHCN winter composite is due to the highly variable nature of the onshore flow from event to event (Smith et al. 1998). During some warm phases, typically the strong events (i.e., 1982 and 1997), onshore flow and heavy precipitation occur in California, while during other warm phases, the onshore flow occurs to the south over the Baja peninsula, resulting in dry conditions over California. The result is a GHCN composite with near-zero anomalies over southern California. Ropelweski and Halpert (1986) initially speculated that warm phase precipitation anomalies along the West Coast may be sensitive to the exact orientation of the midlatitude jet and this is supported by the work of Smith et al. (1998). Montroy (1997) and Montroy et al. (1998) further correlated the location of the anomalous Pacific Ocean warm pool to variability in warm phase precipitation patterns over the eastern United States, and it is likely that shifts in the warm pool position will also impact precipitation over California through modifications in the jet stream patterns.
The pattern of dry anomalies over the northern states and wet conditions along the Gulf Coast during the winter of 1998 is similar to the historical composite. In fact, 97N dry anomalies in Montana, Washington, and Nebraska are larger than most of the historical warm phases studied (Fig. 2). Total winter precipitation at two stations in Nebraska is in the fifth percentile of the warm phase distribution (Fig 1c). Along the Gulf, extreme rainfall was concentrated over Texas and peninsular Florida (Fig. 1b), where some stations received over a 250% increase in rainfall. In both Texas and Florida the heavy precipitation exceeds the composite warm phase conditions (Fig. 1c), with numerous stations having rainfall totals in the 95th percentile of both the warm and neutral phase distributions. Rainfall totals for the winter of the 1998 exceed those of all nine historical warm event winters at four southern plains and five Florida stations (Fig. 2).
A final, notable 1998 winter anomaly pattern is the northward extension of the increased precipitation from the Gulf Coast into the mid-Atlantic states (Fig. 1b). Winter precipitation totals at many stations east of the Appalachians fall into the 95th percentile of the warm phase distribution, and their corresponding 97N anomalies rank as the wettest of the 10 studied warm phases (Fig. 2). This northward extension of wet anomalies may result from an enhancement of the multilevel southwesterly flow identified by Smith et al. (1998) over the entire Gulf of Mexico during a warm phase.
The decay of the warm phase and its associated precipitation anomalies typically occurs in the boreal spring. Sittel (1994b) and Green (1996) found precipitation anomaly patterns for the spring to include a continuation of winter’s dry anomalies in northwestern states and above-neutral precipitation along the Gulf Coast. The GHCN spring composite anomalies agrees with the previous findings; however, our analysis shows the largest percent changes in precipitation occurring over the Southwest (Fig. 3a). Statistically significant decreases in the median of the spring warm phase precipitation are concentrated over the Pacific Northwest, while significant shifts toward increased precipitation are scattered across the lower 48 states.
Precipitation anomaly patterns during spring 1998 are markedly different from the historical spring composite. For example, the Pacific Northwest was wet (Fig. 3b), with many stations having precipitation totals falling in the 95th percentile of both the warm (Fig. 3c) and neutral phase distributions. In fact, seven Pacific Northwest stations received more precipitation in MAM 1998 than occurred in the other nine warm phases (Fig. 4). These anomalous wet conditions are likely a northward extension of the above-neutral precipitation found in California during the winter. The winter wet conditions over California continue in the spring of 1998 (two stations in Nevada received over 300% more precipitation than occurs in neutral years).
Another stark contrast to the historical spring warm phase composite occurs over the Gulf Coast, Texas, and New Mexico. These regions typically have above-neutral precipitation, but instead are dominated by dry anomalies during spring 1998. Three stations in Texas are drier than any of the nine historical warm events (Fig. 4), and many stations in the southern plains have precipitation totals in the fifth percentile of the warm phase gamma distribution (Fig. 3c). This atypical drying was the onset of a disastrous drought that affected most of Texas and the Gulf Coast states into early summer. The forcing behind these atypically dry conditions may be due to the strength of the 1997 warm phase or more likely due to the influence of another large-scale climate pattern (e.g., the North Atlantic oscillation).
When the precipitation anomalies from the 1997 warm event are compared to those from the 1982 warm phase, the only comparable magnitude event in our sample, the patterns were similar with two notable exceptions. Although heavy rains occurred during the winter following the onset of both events, the 1998 winter total were on average 100% larger than the totals for the 1983 winter in southern California, north and central Texas, and the Florida peninsula. In addition, the spring of 1998 was 40%–80% drier than the spring of 1983 in much of the central plains, Texas, and along the Gulf Coast through Florida.
4. 1998 temperature anomalies
Air temperature during a warm phase winter is found to be warmer than neutral phase conditions over the northwest and north-central United States, while the Southwest and portions of the Southeast tend to have cold anomalies (Sittel 1994b; Green 1996). The winter GHCN composite temperature anomalies (Fig. 5a) reveal a similar pattern with the warm anomalies concentrated over Montana and the northern plains. Statistically significant (95% level) differences in the GHCN warm and neutral phase means occur at 12 stations in the northern United States. This warming in the northern plains and Montana is evident in the 97N temperature anomalies (Fig. 5b); however, what is remarkable is the extension of the warm winter anomalies into the Northeast. The resulting 1998 winter temperatures are 3°–5°C warmer than the composite warm phase conditions over much of the Northeast and Great Lakes (Fig. 5c). In fact, the 1998 winter anomalies at the majority of stations east of Wisconsin and Illinois rank as the warmest or second warmest of the 10 studied events (Fig. 6), and are statistically (95%) greater than both historical neutral and warm phase means (Figs. 5b,c).
In the spring, the GHCN composite temperature anomalies (Fig. 7a) show a continuation of warmer than neutral temperatures over the northern states; however, the anomalies are weaker in magnitude and extend farther east than during the previous season (Fig. 5a). This coast to coast warming in the northern states agrees with previously established warm phase conditions (e.g., Sittel 1994b; Green 1996). Much like the winter of 1998, the core of the warming in the spring of 1998 is over the Northeast (Fig. 7b). Many northeastern stations have temperature means that fall into the 95th percentile of both the neutral and warm phase Gaussian distributions, and 12 stations in the Northeast have 97N anomalies that rank as the warmest of the studied warm phases (not shown).
Historical warm phase patterns identify a region of cooler than neutral spring temperature anomalies over much of Texas and the Four Corners states, which is also evident in our GHCN composites (Fig. 7a). Surprisingly, this region of cold anomalies is notably absent in the spring of 1998 (Fig. 7b). Cold anomalies do occur in California and Nevada and are likely linked to increased rainfall in these regions (Fig. 3b). The lack of cold anomalies over Texas is probably due in part (through increased insolation) to the dry anomalies (Fig. 3c) not typically associated with MAM following the peak of a warm phase.
A comparison of the 1998 temperature anomalies to those from the winter and spring following the onset of the 1982 warm phase reveals the eastward shift of the warm anomalies noted earlier (Figs. 5b, 6, and 7b). During the spring of 1998 most stations east of the Rockies are warmer than the spring of 1983. The core of the warmer anomalies is in the Midwest, Great Lakes, and Northeast regions, where 1998 spring temperatures average 2.0°C higher than the spring of 1983.
5. Summary and discussion
Herein we contrast precipitation and temperature anomalies for the winter and spring following the onset of the 1997 warm phase with nine historical warm phases over a 37-yr record. In general, the nine-member composite warm phase anomaly patterns correlate better during the 1998 winter than the spring of 1998. Notable exceptions during the winter of 1998 include the heavy precipitation over California and the Southwest and the extension of the warmer than neutral temperatures into the Northeast.
During the warm phase spring, the Pacific Northwest is typically dry; however, significant wet anomalies were present during MAM 1998. In a similar reversal of patterns, extreme dry anomalies occurred over the southern plains and western Gulf states during spring 1998. Based upon the nine-member composite warm phase, MAM precipitation should be above neutral in these regions. In addition, the composite spring temperature anomalies show the southern plains to be cooler in a warm phase; however, below-neutral temperatures did not occur in 1998. The anomalous drying along the Gulf and warmer temperatures in the southern plains are likely associated with the Bermuda high shifting westward or strengthening.
Finally, the magnitude of the anomalies exceeds the nine historical warm phases from 1951 to 1987 in many cases where the 1998 winter and spring anomaly patterns match the nine-member composite warm phase. For example, although Florida is typically wet during a warm phase, the 1998 winter yielded over 250% more precipitation in Tampa than expected by the composite pattern. Similarly, the 1998 winter temperature anomalies in Minnesota were up to 6°C warmer than the nine-member composite warm phase.
The exact causes for the anomalous precipitation and temperatures during the 1997 warm phase have not been identified, but based upon the works of Smith et al. (1998), Montroy (1997), and others a plausible hypothesis can be devised. The anomalous warming of the eastern tropical Pacific resulted in atypical jet stream patterns during the winter and spring of 1998 over the United States. In many regions, the resulting shifts in storm tracks and dynamic support for precipitation yielded extreme anomalies in precipitation. Additionally, the changes in the jet stream pattern altered the long-wave flow over North America, thereby changing the major air masses that occurred over the United States and modifying seasonal mean temperatures.
At this point, one must consider why the anomalies from the winter and spring of 1998 differ in some regions from the nine-member composite warm phase. First, the nine-member composite will reduce the influence of extreme events and not completely capture the anomaly patterns associated with strong ENSO warm phases. In addition, the nine-member composite will not necessarily represent the typical warm phase anomaly pattern at any one location; it is purely an average of the events included in the composite. A significantly larger sample of warm phase precipitation and temperature anomalies over the United States would be required to determine the typical ENSO signal. With this in mind, we must also consider that some of the 1998 winter and spring anomalies may be unique to the 1997 warm event.
The authors also note that it is unlikely that all the 1998 anomalies were a direct result of the warming in the tropical Pacific Ocean. Interseasonal variations both in the atmosphere over the North Atlantic Ocean (the North Atlantic oscillation; Hurrell 1996) and in the waters of the North Pacific Ocean (Mantau et al. 1997) have been shown to significantly affect surface temperatures over North America. It would be reasonable to expect that varying combinations of the influences of ENSO, the North Atlantic oscillation, and other interseasonal oscillations will result in different anomaly patterns over North America from one ENSO warm phase to the next.
Finally we wish to add a note about decadal-scale variability. In recent years, growing evidence has shown that surface temperatures are affected by oceanic changes on decadal scales (Livezey and Smith 1999). In addition, Mantau et al. (1997) showed the North Atlantic oscillation to affect North American temperatures on a decadal scale. The ENSO cycle has also been shown to have undergone changes in the frequency and intensity of the Pacific warming. From these results, one would expect that seasonal temperature and precipitation anomalies are in part affected by these longer timescale variations. The fact that increased precipitation anomalies in the mid-Atlantic states occurs during the 1998 winter and to some extent in the GHCN composite, but not necessarily in a century-scale composite (Ropelewski and Halpert 1996), may be an indicator of decadal-scale variability in the ENSO response over the United States.
The anomaly patterns for the 1997 warm phase highlight the sometimes contrasting impacts of ENSO warm events. Further research is necessary to quantify these findings in order to more fully understand the variations from one event to the next and to aid their eventual prediction. Accurate ENSO forecasts will only be obtained from improved coupled ocean–atmosphere models that have the ability to predict this event to event variability.
The authors wish to thank Dr. C. F. Ropelewski and the anonymous reviewer for their constructive comments and suggestions. Support was provided by NOAA, Office of Global Programs (Grant NA76GP0521). COAPS receives its base funding from the Physical Oceanography Section of the Office of Naval Research.
Green, P. M., 1996: Regional analysis of Canadian, Alaskan, and Mexican precipitation and temperature anomalies for ENSO impact. COAPS Tech. Rep. 96-6, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL, 104 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.].
Hollander, M., and D. A. Wolfe, 1999: Nonparametric Statistical Methods. John Wiley and Sons, 787 pp.
Hurrell, J. W., 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett.,23, 665–668.
Kiladis, G. N., and H. F. Diaz, 1989: Global climatic anomalies associated with extremes in the Southern Oscillation. J. Climate,2, 1069–1090.
Können, G. P., P. D. Jones, M. H. Kaltofen, and R. S. Allan, 1998: Pre-1866 extensions of the Southern Oscillation Index using early Indonesian and Tahitian meteorological readings. J. Climate,11, 2325–2339.
Livezey, R. E., and T. M. Smith, 1999: Covariability of aspects of North American climate with global sea surface temperatures on interannual to interdecadal timescales. J. Climate,12, 289–302.
Mantau, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc.,78, 1069–1079.
Montroy, D. L., 1997: Linear relation of central and eastern North American precipitation to tropical Pacific sea surface temperature anomalies. J. Climate,10, 541–558.
——, M. B. Richman, and P. J. Lamb, 1998: Observed nonlinearities on monthly teleconnections between tropical Pacific sea surface temperature anomalies and central and eastern North American precipitation. J. Climate,11, 1812–1835.
Ropelewski, C. F., and M. S. Halpert, 1996: Quantifying Southern Oscillation–precipitation relationships. J. Climate,9, 1043–1059.
Shabbar, A., B. Bonsal, and M. Khandekar, 1997: Canadian precipitation patterns associated with the Southern Oscillation. J. Climate,10, 3016–3027.
Sittel, M. C., 1994a: Marginal probabilities of the extremes of ENSO events for temperature and precipitation in the southeastern United States. COAPS Tech. Rep. 94-1, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL, 156 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.].
——, 1994b: Differences in the means of ENSO extremes for maximum temperature and precipitation in the United States. COAPS Tech. Rep. 94-2, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL, 50 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.].
Smith, S. R., P. M. Green, A. P. Leonardi, and J. J. O’Brien, 1998: Role of multiple-level tropospheric circulations in forcing ENSO winter precipitation anomalies. Mon. Wea. Rev.,126, 3102–3116.
Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc.,78, 2771–2777.
Vose, R. S., R. L. Schmoyer, P. M. Steurer, T. C. Peterson, R. Heim, T. R. Karl, and J. K. Eischeid, 1992: The Global Historical Climatology Network: Long-term monthly temperature, precipitation, sea-level pressure, and station pressure data. Oak Ridge National Laboratory, Environmental Science Division Rep. 3912, Oak Ridge, TN, 100 pp. [Available from Oak Ridge National Laboratory, Environmental Science Division, Oak Ridge, TN 37831-6335.].
Wilks, D. S., 1990: Maximum likelihood estimation for the gamma distribution using data containing zeros. J. Climate,3, 1495–1501.
——, 1995: Statistical Methods in the Atmospheric Sciences. Academic Press, 467 pp.
Yarnal, B., and H. Diaz, 1986: Relationships between extremes of the Southern Oscillation and the winter climate of the Anglo-American Pacific Coast. J. Climatol.,6, 197–219.
ENSO warm and neutral phase years based on the JMA SST index. Each year marks the beginning of the ENSO year (e.g., 1951 = October 1951 through September 1952). Note that the onset of warm phase conditions in the equatorial Pacific occurs in the listed years.